Performance Benchmarks

[ashwinrrao]

Hi,
Is there any data available with respect to performance benchmarks of the OTM product?
Benchmarks include infrastructure details, transaction details and more detailed analysis of performance.
Hardware platform being Sun Solaris
Regards,
Ashwin

[chrisplough]

Re: Performance Benchmarks

Ashwin,

There are no published benchmarks regarding OTM and it's ability to handle certain types of transactions. The application is too young and there user-base is to broad (using quite varied functionality) for this to be practical.

However, back and G-Log, I ran much of the application benchmarks and we used a variety of tests to test application performance between releases, patches and platforms:

• VolanoMark - Volano: The Volano Report & Benchmark Tests
  • A great Java benchmark that is highly multithreaded (like OTM) and is very useful in determining relative Java performance between platforms and JDKs.
• Chat - Linux Benchmark Suite Homepage
  • A great C-based benchmark that is highly multi-threaded. Great for determining native application performance on a platform, but not good at determining Java performance. For instance, AIX scores high on chat, since the CPUs and hardware are fantastic, but low on VolanoMark, due to the poor JDK.
• Custom integration (orders, saws, bulk plans, etc) via custom Perl and Python scripts. (Here are MavenWire, we're developing these again in Ruby -- our language of choice.)
• Custom Web UI actions (finders, searches, reports, etc) via custom JMeter scripts. JMeter is a free and highly-configurable web-scripting tool.
  • JMeter - JMeter - Apache JMeter

I can tell you that Solaris is the 3rd best platform for OTM, trailing Linux (1) and Windows (2) in terms of performance. In terms of stability, it's the 2nd best, trailing Linux. Just ensure that you don't use the Sun Niagara CPUs - they are intended for high-threaded, low computational applications and are provide VERY poor performance for OTM.

Thanks,
Chris

[ashwinrrao]

Re: Performance Benchmarks

Hi Chris,
Thanks a lot for this information. This really helps.
Regards,
Ashwin

[chrisplough]

Re: Performance Benchmarks

Just a quick note - I just wrote up a tutorial installing, configuring and running VolanoMark that may be helpful:

Benchmarking Java JVM Performance with VolanoMark - Oracle OTM / G-Log GC3 Community Support

Thanks!
--Chris

[Stefano]

Re: Performance Benchmarks

Hello Chris
We are migrating from 5.0 to 5.5, from windows single box installation to linux OAS + aix DB installation more powerful, after migrating DB, we same data we have DOUBLE TIME to do same operation on new server respect actual one, expeccialy on BUILD MULTISTOP SHIPMENT, we have a SR opened on it from a long time, consider that we have same db, so same rate, same orders....same same...
Do you have any IDEA
Could be that OAS is not so good as BEA?? too new for OTM??
Thanks for help me

[nick.polites]

Re: Performance Benchmarks

Hello Chris
We are migrating from 5.0 to 5.5, from windows single box installation to linux OAS + aix DB installation more powerful, after migrating DB, we same data we have DOUBLE TIME to do same operation on new server respect actual one, expeccialy on BUILD MULTISTOP SHIPMENT, we have a SR opened on it from a long time, consider that we have same db, so same rate, same orders....same same...
Do you have any IDEA
Could be that OAS is not so good as BEA?? too new for OTM??
Thanks for help me

Stefano,

OAS will not perform as well as BEA and the reason is that the JDK that is used (Sun) is much slower at processing than the one used for Weblogic which is BEA's. We don't unfortunately don’t have any benchmarks to prove this as we don't use OAS. The other thing to consider is that OTM was built on Weblogic and has only been on OAS for a few short months now. It will most likely get better in the future but won’t be at the level it is with Weblogic.

Nick

[chrisplough]

Re: Performance Benchmarks

Stephano,

I completely agree with Nick -- matter of fact he manages our Hosting Services and knows as much about OTM performance as I do. The absolute best OTM performance has been on the Linux/JRockit/WebLogic platform -- both because JRockit outperforms the stock Sun 1.4 JDK (sometimes by 200-300%) and because OTM has run on WebLogic for over 8 years.

Instead of just taking our word for it though, I'd recommend running some system benchmarks on your own to verify:
Benchmarking Java JVM Performance with VolanoMark - Oracle OTM / G-Log GC3 Community Support

Finally, here are some of my particular thoughts (rants) around OAS vs WebLogic:
BEA WebLogic vs. Oracle OAS with OTM (G-Log) - Oracle OTM / G-Log GC3 Community Support

--Chris

[Stefano]

Re: Performance Benchmarks

Hello
Thanks Chris and Nick for your answers.

I work on it..

Cheers